Estimating mechanical impedance from hydrophone measurements.

Integrating machine learning or deep learning models into autonomous underwater vehicles often requires extra steps with an accuracy trade-off, and the scarcity of training datasets makes it inconvenient. This work introduces a reduced-order identification approach for thin spherical shells using sparse hydrophone measurements. Because these impedances are unique to a given scatterer, the central hypothesis is that they provide strong identification potential. The method begins by using established techniques to analytically reconstruct the scattered field from hydrophone measurements. From this field, the surface pressures and velocities corresponding to each spherical harmonic are obtained and used to compute the in vacuo mechanical impedance. The method's effectiveness is then evaluated using synthetic data with added noise. Results demonstrate that Modal Mechanical Impedance Estimation can approximate the first two modal mechanical impedances with absolute percentage error less than 10% in the low-frequency range (ka ≤ 2.1) with only 10 hydrophones. Its performance is limited by directivity due to increasing frequency. By enabling efficient computation without any training phase, the proposed method stands out as a promising candidate for real-time and low-energy applications.